Extended Data relating to Main Figure 2/Assessment of ALDH1A1+ cells from primary, treatment-naive glioblastoma. A, Bar plot shows ALDH-bright (ALDHbr) vital cells from two additional paired cases of naive vs. experimental (TMZ→eR) and clinical (cR) relapse. Data as mean ± SD. p values by one-way analysis of variance (ANOVA) with Tukey´s post-hoc test. B, Dotplot shows ALDH1A1 mean fluorescence intensity (MFI) in paired BN46 treatment-naive vs. experimental relapse conditions (in vitro exposure to TMZ (TMZ→eR) or irradiation (RT→eR)). Data are normalized to isotype control, mean ± SD. p values calculated by Kruskal-Wallis test with Dunn's post-hoc test. C, Relative ALDH1A1 expression in -knockdown (shALDH1A1) and -overexpression (Ovx) BN46 cells used for indicated experiment. Data shown as mean ± SD, normalized to their respective controls (ALDH1A1 Ovx to GFP Ctrl | shALDH1A1 to shNT Ctrl). D, Left: Brightfield image of BN46 cells in 96-well plates during monitoring by software-based cell recognition in the limiting dilution assay (NyOne®). An exemplary single cell/well is shown at one day post seeding; representative monoclonal colonies of ALDH1A1-knockdown (sh) and -overexpressing (Ovx) cells at day 16 after seeding. Right: Doubling time estimated at day 16 after seeding (see Methods). Data as mean ± SD, p values calculated by Kruskal-Wallis test with Dunn's post-hoc test. E, Left: Cartoon describes the neurosphere experiments shown in Fig. 2H. Right: Phase contrast microscopic appearance of plated 2° neurosphere and respective immunofluorescence visualization of antibody labeling on neurosphere-derived cells. Neuronal phenotype, β3-tubulin (β3-tub); glial phenotypes, glial fibrillary acidic protein (GFAP). Nuclei exposed with DAPI. Scale bars: left: 100 µM, right: 50 µM. F, Table showing quantification data of 1° and 2° neurosphere generations from (E) in the respective treatment-naive BN46 cells. G, All plated 2° neurospheres from assay (E,F) generated neuronal and glial cell phenotypes.

Extended Data relating to Main Figure 4/Targeting of AKT-driven subclonal progression of ALDH1A1+ cells. A, Collected source data for main Figs. 4B,C,F-H. Left, diagrams: in vitro cell confluence dynamics of paired treatment naive vs. TMZ-related experimental (TMZ→eR; BN46) or clinical (RT/TMZ; BN118, BN123, BN132) relapse patient cells. Monitoring of cell confluence by software-based cell recognition. Assay specified in Fig. 4A. Right, graphs: Readout results from the indicated cases in line, representing source data for Figs. 4B (Cell Confluence), 4C (Cell Viability), and 4H (Apoptosis). Respective assays specified in the main Figure Legends. Data as mean ± SD. p values calculated by pairwise Wilcoxon rank-sum test followed by multiple testing correction using Benjamini-Hochberg method. B, IF of ALDH1A1 and Ki67 in the BN46-donor cell tumor PDX model. Shown is an example from a TMZ+MK2206-treated mouse (OS = 148 days; low levels of cellular co-expression). Scale bar: 20 µm. C, Comparison of overall survival under the investigated treatment conditions (refer to Fig. 4J). Hazard ratio 95% confidence interval estimates and significance levels derived from Cox regression are shown.

Extended Data relating to Main Figure 3/AKT-driven progression of ALDH1A1+ cells in glioblastoma. A, Flow cytometry histograms showing ALDH1A1 expression in the paired treatment-naive vs. clinical relapse patient samples quantified in main Fig. 3A. Isotype controls in gray. B, Flow cytometry histograms of ALDH1A1 derived from treatment-naive and paired experimental relapse (TMZ→eR) BN46 cells. Data quantification in Supplementary Fig. S2B. C, Cartoon illustrating course of neurosphere experiments, quantified in main Fig. 3B and in Supplementary Fig. S3D, applying treatment-naive and paired experimental and clinical relapse patient cells. Right: Exemplary qRT-PCR data presenting knockdown efficacy of the siRNA approach in respective cells, normalized to siNT control. Mean of duplicates ± SD. D, Source data for Fig. 3B, and for Supplementary Fig. S3E. Dotplots show percent neurosphere-forming cells estimated from neurospheres at 12 days after seeding. Paired patient cell analysis, evaluating 1° and 2° neurospheres of indicated knockdown (siALDH1A1) and respective control (siNT) cell samples. Data shown as mean ± SD. E, Neurosphere assay, similar to main Fig. 3B on paired treatment-naive vs. experimental relapse (TMZ→eR) BN46 cells. Source data in Supplementary Fig. S3D. Data as mean ± SD. F, Protein patterns, similar to Fig. 3D. on paired cells from 2 additional discovery cohort samples. G, Flow cytometry histograms depicting pAKT(Ser473) expression in the paired treatment-naive vs. clinical relapse patient samples quantified in main Fig. 3E. Isotype controls in gray. H, Flow cytometry histograms of pAKT(Ser473) and, right, quantification of data derived from treatment-naive and paired experimental relapse (TMZ→eR) BN46 cells. Data represent mean fluorescence intensities (MFI), ± SD, normalized to isotype control (gray). I, Representative source data for Fig. 3F. Flow cytometry profiles of ALDH1A1/pAKT(Ser473)-labeled, paired treatment-naive vs. clinical relapse cell samples. J, Flow cytometry profiles of of ALDH1A1/pAKT(Ser473)-labeled, paired treatment-naive vs. experimental relapse conditions of BN46 cells (in vitro exposure to TMZ (TMZ→eR) or irradiation (RT→eR)).

Validation cohort, patient flow diagram. Patient enrollment and final inclusion of FFPE tissue samples in this study is shown. n = 3 patients were excluded from enrollment for reasons listed. The validation cohort comprised paired tissue samples from a total of n = 38 patients. Detailed characteristics in Supplementary Table S4. Note, the separate discovery cohort of paired vital cells is detailed in Supplementary Table S1. The GLASS cohort was used as an additional, external reference comprising data on n=37 patients extracted from (10).

Supplementary Table S1. Discovery cohort of vital cells; Supplementary Table S2. Validation cohort/FFPE tissue samples; Supplementary Table S3. Differentially regulated genes; Supplementary Table S4. Selected gene variants compared between the paired treatment-naive vs. relapse cell samples.

The ecosystem of brain tumors is considered immunosuppressed, but our current knowledge may be incomplete. Here we analyzed clinical cell and tissue specimens derived from patients presenting with glioblastoma or nonmalignant intracranial disease to report that the cranial bone (CB) marrow, in juxtaposition to treatment-naive glioblastoma tumors, harbors active lymphoid populations at the time of initial diagnosis. Clinical and anatomical imaging, single-cell molecular and immune cell profiling and quantification of tumor reactivity identified CD8+ T cell clonotypes in the CB that were also found in the tumor. These were characterized by acute and durable antitumor response rooted in the entire T cell developmental spectrum. In contrast to distal bone marrow, the CB niche proximal to the tumor showed increased frequencies of tumor-reactive CD8+ effector types expressing the lymphoid egress marker S1PR1. In line with this, cranial enhancement of CXCR4 radiolabel may serve as a surrogate marker indicating focal association with improved progression-free survival. The data of this study advocate preservation and further exploitation of these cranioencephalic units for the clinical care of glioblastoma. Analyses of tumor and bone marrow tissue from patients with glioblastoma demonstrate the presence of extracerebral niches that contained tumor-reactive and memory T cell subsets, including early stem-like phenotypes and stages, indicating antitumor CD8+ T cell differentiation in cranial bone marrow.

Abstract Emerging evidence supports the notion that phenotypic plasticity contributes to disease progression and drug resistance in malignant glioma. We have recently described a rare population of ALDH1A1+ tumor cells in newly diagnosed, treatment-naive glioblastoma that can adapt to the exposure of the standard chemotherapeutic temozolomide (TMZ). Their subclonal growth leads to AKT-driven, TMZ-resistant cellular hierarchies. Accumulation of ALDH1A1+/pAKT+ cells can therefore be noted subsequent to TMZ in patient relapse tissue and in PDX models of disease. We have also shown that this series of events requires a sequential targeting approach where these subclones are allowed to enrich under TMZ and, only in a second step, are treated with AKT inhibitors. This “enrich and kill” strategy doubles the TMZ-based survival benefit in preclinical PDX models (Kebir et al., Clin Cancer Res 2023). Here, we took advantage of short-term expanded patient-derived cell cultures, which allow the study of early stages of this type of adaptive plasticity in controlled conditions. Specifically, we characterized the human lysine-specific demethylase 5B (KDM5B) as a prospective indicator for subclonal expansion in rare cells under first-time exposure to TMZ. We used genetic reporters, pharmacological interference, CHIP-seq analysis and cellular barcoding to investigate the dynamics of KDM5B expression and the related intracellular cross-signaling. We monitored patient cell samples over prolonged periods of time, and we found that KDM5Bhigh treatment-naive glioblastoma cells preferentially contribute to the dynamics of ALDH1A1 subclones and drug resistance under the influence of TMZ. These findings may lay ground for the development of biomarker-assisted clinical trials. This work is supported by DFG/GRC-CRU337/2 (proj#405344257).

AbstractPurpose:Therapy resistance and fatal disease progression in glioblastoma are thought to result from the dynamics of intra-tumor heterogeneity. This study aimed at identifying and molecularly targeting tumor cells that can survive, adapt, and subclonally expand under primary therapy.Experimental Design:To identify candidate markers and to experimentally access dynamics of subclonal progression in glioblastoma, we established a discovery cohort of paired vital cell samples obtained before and after primary therapy. We further used two independent validation cohorts of paired clinical tissues to test our findings. Follow-up preclinical treatment strategies were evaluated in patient-derived xenografts.Results:We describe, in clinical samples, an archetype of rare ALDH1A1+ tumor cells that enrich and acquire AKT-mediated drug resistance in response to standard-of-care temozolomide (TMZ). Importantly, we observe that drug resistance of ALDH1A1+ cells is not intrinsic, but rather an adaptive mechanism emerging exclusively after TMZ treatment. In patient cells and xenograft models of disease, we recapitulate the enrichment of ALDH1A1+ cells under the influence of TMZ. We demonstrate that their subclonal progression is AKT-driven and can be interfered with by well-timed sequential rather than simultaneous antitumor combination strategy.Conclusions:Drug-resistant ALDH1A1+/pAKT+ subclones accumulate in patient tissues upon adaptation to TMZ therapy. These subclones may therefore represent a dynamic target in glioblastoma. Our study proposes the combination of TMZ and AKT inhibitors in a sequential treatment schedule as a rationale for future clinical investigation.

Adaptive plasticity to the standard chemotherapeutic temozolomide (TMZ) leads to glioblastoma progression. Here, we examine early stages of this process in patient-derived cellular models, exposing the human lysine-specific demethylase 5B (KDM5B) as a prospective indicator for subclonal expansion. By integration of a reporter, we show its preferential activity in rare, stem-like ALDH1A1+ cells, immediately increasing expression upon TMZ exposure. Naive, genetically unmodified KDM5Bhigh cells phosphorylate AKT (pAKT) and act as slow-cycling persisters under TMZ. Knockdown of KDM5B reverses pAKT levels, simultaneously increasing PTEN expression and TMZ sensitivity. Pharmacological inhibition of PTEN rescues the effect. Interference with KDM5B subsequent to TMZ decreases cellular vitality, and clonal tracing with DNA barcoding demonstrates high individual levels of KDM5B to predict subclonal expansion already before TMZ exposure. Thus, KDM5Bhigh treatment-naive cells preferentially contribute to the dynamics of drug resistance under TMZ. These findings may serve as a cornerstone for future biomarker-assisted clinical trials.

In the version of the article published, the author list is not accurate. Igor Cima and Min-Han Tan should have been authors, appearing after Mark Wong in the author list, while Paul Jongjoon Choi should not have been listed as an author. Igor Cima and Min-Han Tan both have the affiliation Institute of Bioengineering and Nanotechnology, Singapore, Singapore, and their contributions should have been noted in the Author Contributions section as "I.C. preprocessed Primary Cell Atlas data with inputs from M.-H.T." The following description of the contribution of Paul Jongjoon Choi should not have appeared: "P.J.C. supported the smFISH experiments." In the 'RCA: global panel' section of the Online Methods, the following sentence should have appeared as the second sentence, "An expression atlas of human primary cells (the Primary Cell Atlas) was preprocessed similarly to in ref. 55," with new reference 55 (Cima, I. et al. Tumor-derived circulating endothelial cell clusters in colorectal cancer. Science Transl. Med. 8, 345ra89, 2016).

Immune evasion is indispensable for cancer initiation and progression, although its underlying mechanisms in pancreatic ductal adenocarcinoma (PDAC) are not fully known. Here, we characterize the function of tumor-derived PGRN in promoting immune evasion in primary PDAC. Tumor- but not macrophage-derived PGRN is associated with poor overall survival in PDAC. Multiplex immunohistochemistry shows low MHC class I (MHCI) expression and lack of CD8+ T cell infiltration in PGRN-high tumors. Inhibition of PGRN abrogates autophagy-dependent MHCI degradation and restores MHCI expression on PDAC cells. Antibody-based blockade of PGRN in a PDAC mouse model remarkably decelerates tumor initiation and progression. Notably, tumors expressing LCMV-gp33 as a model antigen are sensitized to gp33-TCR transgenic T cell-mediated cytotoxicity upon PGRN blockade. Overall, our study shows a crucial function of tumor-derived PGRN in regulating immunogenicity of primary PDAC.

Brain tumors are typically immunosuppressive and refractory to immunotherapies for reasons that remain poorly understood. The unbiased profiling of immune cell types in the tumor microenvironment may reveal immunologic networks affecting therapy and course of disease. Here we identify and validate the presence of hematopoietic stem and progenitor cells (HSPCs) within glioblastoma tissues. Furthermore, we demonstrate a positive link of tumor-associated HSPCs with malignant and immunosuppressive phenotypes. Compared to the medullary hematopoietic compartment, tumor-associated HSPCs contain a higher fraction of immunophenotypically and transcriptomically immature, CD38- cells, such as hematopoietic stem cells and multipotent progenitors, express genes related to glioblastoma progression and display signatures of active cell cycle phases. When cultured ex vivo, tumor-associated HSPCs form myeloid colonies, suggesting potential in situ myelopoiesis. In experimental models, HSPCs promote tumor cell proliferation, expression of the immune checkpoint PD-L1 and secretion of tumor promoting cytokines such as IL-6, IL-8 and CCL2, indicating concomitant support of both malignancy and immunosuppression. In patients, the amount of tumor-associated HSPCs in tumor tissues is prognostic for patient survival and correlates with immunosuppressive phenotypes. These findings identify an important element in the complex landscape of glioblastoma that may serve as a target for brain tumor immunotherapies.

Early detection of cancer holds high promise for reducing cancer-related mortality. Detection of circulating tumor-specific nucleic acids holds promise, but sensitivity and specificity issues remain with current technology. We studied cell-free RNA (cfRNA) in patients with non-small cell lung cancer (NSCLC; n = 56 stage IV, n = 39 stages I-III), pancreatic cancer (PDAC, n = 20 stage III), malignant melanoma (MM, n = 12 stage III-IV), urothelial bladder cancer (UBC, n = 22 stage II and IV), and 65 healthy controls by means of next generation sequencing (NGS) and real-time droplet digital PCR (RT-ddPCR). We identified 192 overlapping upregulated transcripts in NSCLC and PDAC by NGS, more than 90% of which were noncoding. Previously reported transcripts (e.g., HOTAIRM1) were identified. Plasma cfRNA transcript levels of POU6F2-AS2 discriminated NSCLC from healthy donors (AUC = 0.82 and 0.76 for stages IV and I–III, respectively) and significantly associated (p = 0.017) with the established tumor marker Cyfra 21-1. cfRNA yield and POU6F2-AS transcript abundance discriminated PDAC patients from healthy donors (AUC = 1.0). POU6F2-AS2 transcript was significantly higher in MM (p = 0.044). In summary, our findings support further validation of cfRNA detection by RT-ddPCR as a biomarker for early detection of solid cancers.

Deformability is shown to correlate with the invasiveness and metastasis of cancer cells. Recent studies suggest epithelial-to-mesenchymal transition (EMT) might enable cancer metastasis. However, the correlation of EMT with cancer cell deformability has not been well elucidated. Cellular deformability could also help evaluate the drug response of cancer cells. Here, we combine hydrodynamic stretching and microsieve filtration to study cellular deformability in several cellular models. Hydrodynamic stretching uses extensional flow to rapidly quantify cellular deformability and size with high throughput at the single cell level. Microsieve filtration can rapidly estimate relative deformability in cellular populations. We show that colorectal cancer cell line RKO with the mesenchymal-like feature is more flexible than the epithelial-like HCT116. In another model, the breast epithelial cells MCF10A with deletion of the TP53 gene are also significantly more deformable compared to their isogenic wildtype counterpart, indicating a potential genetic link to cellular deformability. We also find that the drug docetaxel leads to an increase in the size of A549 lung cancer cells. The ability to associate mechanical properties of cancer cells with their phenotypes and genetics using single cell hydrodynamic stretching or the microsieve may help to deepen our understanding of the basic properties of cancer progression.

Early cancer detection, its monitoring, and therapeutical prediction are highly valuable, though extremely challenging targets in oncology. Significant progress has been made recently, resulting in a group of devices and techniques that are now capable of successfully detecting, interpreting, and monitoring cancer biomarkers in body fluids. Precise information about malignancies can be obtained from liquid biopsies by isolating and analyzing circulating tumor cells (CTCs) or nucleic acids, tumor-derived vesicles or proteins, and metabolites. The current work provides a general overview of the latest on-chip technological developments for cancer liquid biopsy. Current challenges for their translation and their application in various clinical settings are discussed. Microfluidic solutions for each set of biomarkers are compared, and a global overview of the major trends and ongoing research challenges is given. A detailed analysis of the microfluidic isolation of CTCs with recent efforts that aimed at increasing purity and capture efficiency is provided as well. Although CTCs have been the focus of a vast microfluidic research effort as the key element for obtaining relevant information, important clinical insights can also be achieved from alternative biomarkers, such as classical protein biomarkers, exosomes, or circulating-free nucleic acids. Finally, while most work has been devoted to the analysis of blood-based biomarkers, we highlight the less explored potential of urine as an ideal source of molecular cancer biomarkers for point-of-care lab-on-chip devices.

Studies on circulating tumor cells (CTCs) have largely focused on platform development and CTC enumeration rather than on the genomic characterization of CTCs. To address this, we performed targeted sequencing of CTCs of colorectal cancer patients and compared the mutations with the matched primary tumors. We collected preoperative blood and matched primary tumor samples from 48 colorectal cancer patients. CTCs were isolated using a label-free microfiltration device on a silicon microsieve. Upon whole genome amplification, we performed amplicon-based targeted sequencing on a panel of 39 druggable and frequently mutated genes on both CTCs and fresh-frozen tumor samples. We developed an analysis pipeline to minimize false-positive detection of somatic mutations in amplified DNA. In 60% of the CTC-enriched blood samples, we detected primary tumor matching mutations. We found a significant positive correlation between the allele frequencies of somatic mutations detected in CTCs and abnormal CEA serum level. Strikingly, we found driver mutations and amplifications in cancer and druggable genes such as APC, KRAS, TP53, ERBB3, FBXW7 and ERBB2. In addition, we found that CTCs carried mutation signatures that resembled the signatures of their primary tumors. Cumulatively, our study defined genetic signatures and somatic mutation frequency of colorectal CTCs. The identification of druggable mutations in CTCs of preoperative colorectal cancer patients could lead to more timely and focused therapeutic interventions.

Breast fibroepithelial lesions are biphasic tumors and include fibroadenomas and phyllodes tumors. Preoperative distinction between fibroadenomas and phyllodes tumors is pivotal to clinical management. Fibroadenomas are clinically benign while phyllodes tumors are more unpredictable in biological behavior, with potential for recurrence. Differentiating the tumors may be challenging when they have overlapping clinical and histological features especially on core biopsies. Current molecular and immunohistochemical techniques have a limited role in the diagnosis of breast fibroepithelial lesions. We aimed to develop a practical molecular test to aid in distinguishing fibroadenomas from phyllodes tumors in the pre-operative setting.

Clusters of tumor cells are often observed in the blood of cancer patients. These structures have been described as malignant entities for more than 50 years, although their comprehensive characterization is lacking. Contrary to current consensus, we demonstrate that a discrete population of circulating cell clusters isolated from the blood of colorectal cancer patients are not cancerous but consist of tumor-derived endothelial cells. These clusters express both epithelial and mesenchymal markers, consistent with previous reports on circulating tumor cell (CTC) phenotyping. However, unlike CTCs, they do not mirror the genetic variations of matched tumors. Transcriptomic analysis of single clusters revealed that these structures exhibit an endothelial phenotype and can be traced back to the tumor endothelium. Further results show that tumor-derived endothelial clusters do not form by coagulation or by outgrowth of single circulating endothelial cells, supporting a direct release of clusters from the tumor vasculature. The isolation and enumeration of these benign clusters distinguished healthy volunteers from treatment-naïve as well as pathological early-stage (≤IIA) colorectal cancer patients with high accuracy, suggesting that tumor-derived circulating endothelial cell clusters could be used as a means of noninvasive screening for colorectal cancer. In contrast to CTCs, tumor-derived endothelial cell clusters may also provide important information about the underlying tumor vasculature at the time of diagnosis, during treatment, and throughout the course of the disease.

Characterization of genetic alterations in tumor biopsies serves as useful biomarkers in prognosis and treatment management. Circulating tumor cells (CTCs) obtained non‐invasively from peripheral blood could serve as a tumor proxy. Using a label‐free CTC enrichment strategy that we have established, we aimed to develop sensitive assays for qualitative assessment of tumor genotype in patients. Blood consecutively obtained from 44 patients with local and advanced colorectal cancer and 18 healthy donors were enriched for CTCs using a size‐based microsieve technology. To screen for CTC mutations, we established high‐resolution melt (HRM) and allele‐specific PCR (ASPCR) KRAS‐codon 12/13‐ and BRAF‐codon 600‐ specific assays, and compared the performance with pyrosequencing and Sanger sequencing. For each patient, the resulting CTC genotypes were compared with matched tumor and normal tissues. Both HRM and ASPCR could detect as low as 1.25% KRAS‐ or BRAF‐mutant alleles. HRM detected 14/44 (31.8%) patients with KRAS mutation in CTCs and 5/44 (11.3%) patients having BRAF mutation in CTCs. ASPCR detected KRAS and BRAF mutations in CTCs of 10/44 (22.7%) and 1/44 (2.3%) patients respectively. There was an increased detection of mutation in blood using these two methods. Comparing tumor tissues and CTCs mutation status using HRM, we observed 84.1% concordance in KRAS genotype (p = 0.000129, Fishers' exact test; OR = 38.7, 95% CI = 4.05–369) and 90.9% (p = 0.174) concordance in BRAF genotype. Our results demonstrate that CTC enrichment, coupled with sensitive mutation detection methods, may allow rapid, sensitive and non‐invasive assessment of tumor genotype.